Browse / Software development / clankbrain
clankbrain
YehudaFrankel
Memory + workflows for Claude Code that compound across sessions
Open YehudaFrankel/clankbrain → Does not match the four rules.
ai-toolsclaudeclaude-codecontext-engineeringdeveloper-toolsllmmemoryproductivitypythonworkflow
Works with any agent
ships AGENTS.md, the cross-tool instruction file.
In its own words
Every session, your coding agent wakes up a stranger.
You re-explain the stack. Re-describe what you were building. Re-establish why certain approaches won't work here. Then it fumbles through the same debug flow, the same code review, the same refactor — because it has no structured way to do any of it.
Clankbrain fixes both. It gives Claude Code a living memory — decisions, mistakes, patterns, lessons that compound across sessions — and 37 built-in workflows for every development task: debugging, code review, refactoring, frontend design, security audits, and more.
Session 1: the agent learns your stack. Session 10: it knows every mistake you've made, every approach you've rejected, every shortcut that works in your codebase. Session 20: you stop explaining. You just build.
Two commands. That's the whole interface.
New here? After install, type tour for a 5-minute interactive walkthrough. Or follow FIRST5.md — a guided path through your first 5 sessions, one per real work block. Session 5 is when the compounding becomes visible.
Install
Or if you cloned the repo directly:
Five questions. Creates your memory directory, hooks config, and starter files. Done in 30 seconds.
If anything goes wrong, the installer tells you exactly what's broken and how to fix it (no Python tracebacks).
No API keys. No background service. No database. Zero pip dependencies — stdlib Python only. Requires: Claude Code or Codex + Python 3.7+
Quoting 11% of the README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.